#microformats 2018-02-18

2018-02-18 UTC
tantek, [cleverdevil] and KartikPrabhu joined the channel
#
KartikPrabhu updated mf2py for the link[title] and video[poster] parsing for p-* and u-*
tantek joined the channel
#
tantek KartikPrabhu++
#
Loqi kartikprabhu has 5 karma in this channel (168 overall)
#
KartikPrabhu now we need to figure out how to incorporate these changes into the actual mf2py code base
#
tantek still think push to upstream repo is the way
#
tantek and we can attempt to contact tommorris e.g. on email
#
tantek to move the repo
#
KartikPrabhu I did push the new p-* and u-* rules
#
KartikPrabhu for the implied name parsing I am waiting to see if it makes into the spec
#
tantek that's reasonable. having at least one parser demonstrate implementability is a condition of making the spec change
#
tantek would be nice to hear from snarfed about if it works for his use-case
#
KartikPrabhu yes
[mrkrndvs] and tantek joined the channel
#
KartikPrabhu !tell [kevinmarks]: would be awesome if you could test https://github.com/kartikprabhu/mf2py/tree/implied-name-fix on the examples in https://github.com/microformats/microformats2-parsing/issues/6#issuecomment-366473390 for some reason my local install works correctly but not the installation on my server!
#
KartikPrabhu !tell [kevinmarks]: would be awesome if you could test https://github.com/kartikprabhu/mf2py/tree/implied-name-fix on the examples in https://github.com/microformats/microformats2-parsing/issues/6#issuecomment-366473390 for some reason my local install works correctly but not the installation on my server!
#
Loqi Ok, I'll tell them that when I see them next
#
KartikPrabhu and anyone else too ^
[eddie], [kevinmarks], [tantek], [mrkrndvs], nitot, [miklb], iwaim___, [jeremycherfas], 5EXAAPICA, barpthewire, tantek, [cleverdevil] and [squorch] joined the channel
#
KartikPrabhu [kevinmarks]: would appreciate your help/insight on this https://github.com/kartikprabhu/mf2py/issues/58#issuecomment-366533803
#
Loqi [kartikprabhu] The output depends on which html parser is being used internally. cc @kevinmarks In the example below `html5lib` gives a `name` and `url` for the `h-enty` but `html.parser` does not (which is the intended output). # Example ## HTML ``` html...
#
Zegnat html5lib doesn’t even give a correct implied name there, if one goes by the old rules, KartikPrabhu.
#
Zegnat Also, what’s up with that double url property?
#
KartikPrabhu Zegnat: yeah not sure what is going on
#
Zegnat Not to derail you, sorry :P
#
KartikPrabhu sorry updated comment. the double url is only in html5lib not in html.parser
#
KartikPrabhu html5lib is supposed to be better but has some funny quirks
#
KartikPrabhu html5lib works with the old rules but seems to be doing funny things with the new ones!
#
KartikPrabhu html.parser works with both rules
#
KartikPrabhu so I am very confused since the mf2py code is the same but the parsers are different
#
Zegnat Do they both output DOM with a common API, or how do you access their output?
#
Zegnat is desperately fighting the urge to learn Python just to be part of the conversation
#
[kevinmarks] Beautiful soup is on top of them both iirc.
#
sknebel yes
KartikPrabhu joined the channel
#
sknebel mf2ßy only tells BS which one to use
#
KartikPrabhu Zegnat: this is happening because the <a>-tag in your 3rd example is not closed!
#
KartikPrabhu and html5lib is doing some crazy stuff with it!
#
Zegnat Oh, I’ll double check that in a second KartikPrabhu!
#
sknebel yeah, that seems to be it
#
[kevinmarks] S/crazy/spec compliant/
#
sknebel yes
#
sknebel the other examples seem like correct html
#
Zegnat It is really interesting. Per spec, I thought it was supposed to close that A prior to the </p>
#
Zegnat Either way, my fault, sorry
#
KartikPrabhu [kevinmarks]: https://github.com/kartikprabhu/mf2py/issues/58#issuecomment-366536321 see the corrected tree outputs
#
Loqi [kartikprabhu] @kevinmarks I am going to disagree in this case. Taking the same example # HTML ``` html <article class="h-entry"> <div class="u-like-of h-cite"> <p>I really like <a class="p-name u-url" href="http://microformats.org/">Microformats</p> ...
#
KartikPrabhu I have no idea why html5libis supposed to be the "correct" spec compliant one
#
KartikPrabhu no browser does it I think
#
KartikPrabhu Zegnat: yes that is what I thought too
#
sknebel KartikPrabhu: try it in a browser
#
sknebel Chrome shows me the exact same DOM
#
Zegnat I have fixed the HTML in the original microformats issue
#
KartikPrabhu sknebel: with the repeated <a> ?
#
sknebel I have to correct myself, Chrome repeats it 3 times
#
Zegnat 4 A elements total in Firefox
#
Zegnat Interesssssting. Not what I expected.
#
KartikPrabhu yeah FF has 4 elements <a>
#
KartikPrabhu wow!
#
Zegnat Might be spec compliant though, in which case html5lib is completely right in what it did
#
[kevinmarks] I'm not sure why the spec would expect that, but html5lib is very compatible with what browsers do.
#
KartikPrabhu [kevinmarks]: here it depends on which browser since CHrome and FF do different things
#
KartikPrabhu :P
#
sknebel nah, I didn't exand one element - it's 4 in chrome too
#
KartikPrabhu oh ok
#
KartikPrabhu still differnet from the html5lib output
#
sknebel yeah
#
sknebel I guess this happens because it tries to wrap the <a> around elements that can't be in an <a>, so it gets split into <a>s around the contents of these elements?
#
KartikPrabhu <shrug> who knows!
#
KartikPrabhu so I guess html5lib is the correct one
#
KartikPrabhu or "more" correct
#
sknebel or at least not "more wrong"...
#
sknebel ugh
#
KartikPrabhu I wonder how other parsers handle this one
#
KartikPrabhu mf2 parsers
#
Zegnat Completely dependent on the error handling of the HTML parser they use. I don’t think any parser does the HTML parsing themselves
#
KartikPrabhu Zegnat: yes but it would be good to see the differences since the mf2 parsed output depends on that
#
KartikPrabhu Zegnat: with the corrected example mf2py with html5ib also works
#
KartikPrabhu Zegnat: try it here https://kartikprabhu.com/connection/mfparser
[chrisaldrich] joined the channel
#
Zegnat I am not sure how it helps anything to know how faulty HTML is being “corrected” by different parsers though. Even if it makes the final mf2 wrong, it will always be an upstream error :/
#
[kevinmarks] I expect that the js one that uses browser dom would match. Php will depend on lxml doing its own thing.
#
[kevinmarks] I am wondering if we should force html5lib in mf2py
#
[kevinmarks] It is a bit slower
#
KartikPrabhu [kevinmarks]: could you test that branch with lxml?
#
KartikPrabhu I am not sure how to install a C lib in virtual env
#
sknebel BeautifulSoup picks the best parser it has by default, so it'll use html5lib if available
#
KartikPrabhu sknebel: mf2py has the option to specify a parser
#
sknebel that too. but if you leave that empty, BS picks
#
KartikPrabhu which it basically passes on the BS
#
KartikPrabhu yes
#
sknebel hm, given that html5lib is python it should be installable everywhere.
#
sknebel I think snarfed used other parsers somewhere due to performance?
#
sknebel I guess a step inbetween would be to explicitly make html5lib the default and require the user to override if they want something else?
#
KartikPrabhu sknebel: yeah that seems reasonable. html5lib default unless user specified. if html5lib is not installed defer to BS defaults
#
sknebel that's basically what's currently there, since BS will use html5lib by default if it is there
#
KartikPrabhu no I think it goes for lxml first since that is faster
#
KartikPrabhu is finding docs for that^
#
sknebel oh, right
#
sknebel sorry
#
sknebel it's not finding lxml here
#
KartikPrabhu https://www.crummy.com/software/BeautifulSoup/bs4/doc/#specifying-the-parser-to-use
#
KartikPrabhu lxml > html5lib > html.parser
#
sknebel ah, used the wrong python version. sorry for the confusion
#
KartikPrabhu now check lxml
#
KartikPrabhu ok updated with all trees https://github.com/kartikprabhu/mf2py/issues/58#issuecomment-366536321
#
Loqi [kartikprabhu] @kevinmarks ~~I am going to disagree in this case.~~ Taking the same example # HTML ``` html <article class="h-entry"> <div class="u-like-of h-cite"> <p>I really like <a class="p-name u-url" href="http://microformats.org/">Microformats<...
#
sknebel KartikPrabhu++
#
Loqi kartikprabhu has 6 karma in this channel (169 overall)
#
KartikPrabhu vote https://github.com/kartikprabhu/mf2py/issues/59
#
Loqi [kartikprabhu] #59 Use html5lib by default
[eddie], [cleverdevil] and [kevinmarks] joined the channel
#
KartikPrabhu ok whole new branch of mf2py with default html5lib and handling of redirect URLS along wth implied-name-fixes https://github.com/kartikprabhu/mf2py/tree/parsing-into-bs4
#
KartikPrabhu enough mf2py for this weekend (for now)
#
Zegnat Woop for active parser development!
#
Zegnat KartikPrabhu++
#
Loqi kartikprabhu has 7 karma in this channel (170 overall)
GWG, [pfefferle], [miklb], KartikPrabhu and [cleverdevil] joined the channel